A Method for Detecting an Ironic Expression

Abstract

We designed a method for detecting ironic utterances. The system is based on the fact that ironies generally satisfy 4 conditions : (1) the situation has a negative "personal evaluation value" (PEV), (2) the surface meaning of the utterance has a positive PEV, (3) the situation and the utterance are opposite in meaning, and (4) the utterance includes "ironic markers". The system, whose input is the situation and the utterance, consists of 4 processing steps. First, the system determines the PEV of the input. Second, the System checks whether the input satisfies conditons (l) and (2). Third, the system compares the opposite relation between the situation and the utterance. And fourth, the system inspects the sentence particle in the utterance. The output is the calculated "degree of irony". The proposed method contributes to the improvement of the understanding system of connotation in natural language.